Conventionally, an information extraction apparatus is developed to extract numerical data associated with parts-of-speech such as noun upon performing morphological analysis on text data (see, Patent Document 1 for example). The conventional apparatus cuts out the text data one sentence at a time and extracts sentences having numerical values. A judgment is then made for sentence modification and phrases associated with numerical values are extracted.    Patent Document 1: Japanese Patent Application Publication No. 2005-149359